Machine Learning in Differentiating Gliomas from Primary CNS Lymphomas: A Systematic Review, Reporting Quality, and Risk of Bias Assessment.

BACKGROUND: Differentiating gliomas and primary CNS lymphoma represents a diagnostic challenge with important therapeutic ramifications. Biopsy is the preferred method of diagnosis, while MR imaging in conjunction with machine learning has shown promising results in differentiating these tumors. PURPOSE: Our aim was to evaluate the quality of reporting and risk of bias, assess data bases with which the machine learning classification algorithms were developed, the algorithms themselves, and their performance. DATA SOURCES: Ovid EMBASE, Ovid MEDLINE, Cochrane Central Register of Controlled Trials, and the Web of Science Core Collection were searched according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. STUDY SELECTION: From 11,727 studies, 23 peer-reviewed studies used machine learning to differentiate primary CNS lymphoma from gliomas in 2276 patients. DATA ANALYSIS: Characteristics of data sets and machine learning algorithms were extracted. A meta-analysis on a subset of studies was performed. Reporting quality and risk of bias were assessed using the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) and Prediction Model Study Risk Of Bias Assessment Tool. DATA SYNTHESIS: The highest area under the receiver operating characteristic curve (0.961) and accuracy (91.2%) in external validation were achieved by logistic regression and support vector machines models using conventional radiomic features. Meta-analysis of machine learning classifiers using these features yielded a mean area under the receiver operating characteristic curve of 0.944 (95% CI, 0.898-0.99). The median TRIPOD score was 51.7%. The risk of bias was high for 16 studies. LIMITATIONS: Exclusion of abstracts decreased the sensitivity in evaluating all published studies. Meta-analysis had high heterogeneity. CONCLUSIONS: Machine learning-based methods of differentiating primary CNS lymphoma from gliomas have shown great potential, but most studies lack large, balanced data sets and external validation. Assessment of the studies identified multiple deficiencies in reporting quality and risk of bias. These factors reduce the generalizability and reproducibility of the findings.

Imaging of Neuromodulation and Surgical Interventions for Epilepsy.

About one-third of epilepsy cases are refractory to medical therapy. During the past decades, the availability of surgical epilepsy interventions has substantially increased as therapeutic options for this group of patients. A wide range of surgical interventions and electrophysiologic neuromodulation techniques are available, including lesional resection, lobar resection, thermoablation, disconnection, multiple subpial transections, vagus nerve stimulation, responsive neurostimulation, and deep brain stimulation. The indications and imaging features of potential complications of the newer surgical interventions may not be widely appreciated, particularly if practitioners are not associated with comprehensive epilepsy centers. In this article, we review a wide range of invasive epilepsy treatment modalities with a particular focus on their postoperative imaging findings and complications. A state-of-the-art treatment algorithm provides context for imaging findings by helping the reader understand how a particular invasive treatment decision is made.

Brainstem melanomas presenting as a cavernous malformation.

BACKGROUND: Melanoma lesions in the brainstem can be difficult to distinguish radiographically and clinically from cavernous malformations. However, the treatment modalities and clinical course of these two diseases differ considerably. We report two cases of melanoma presenting as brainstem hemorrhages. CASE DESCRIPTION: A 69-year-old male was found to have a hemorrhagic lesion of the right dorsal midbrain. After a repeat hemorrhage, the lesion was resected and found to be hyperchromatic. Nonetheless, the patient suffered rebleeding and died 3 months later. A 62-year-old female was similarly found to have an acute pontine hemorrhage. After resection of the lesion, she underwent whole-brain radiation therapy but ultimately died 5.5 months later. The histopathology of both lesions was consistent with melanoma. CONCLUSIONS: Melanoma in the brainstem can mimic cavernous malformations. While management of these lesions includes stereotactic radiosurgery, whole-brain radiation, and surgical resection, metastatic brainstem melanoma follows an aggressive clinical course with a poor prognosis.

Choroid plexus changes after temporal lobectomy.

BACKGROUND AND PURPOSE: Postoperative contrast-enhanced MR imaging of the brain is routinely used when evaluating for residual or recurrent brain tumor. It is imperative to be aware of morphologic changes and imaging features that typically occur in response to surgical manipulation at the postoperative site to avoid misinterpretation of imaging findings. Our purpose was to determine normal postoperative changes and alterations in the choroid plexus among patients who had undergone temporal lobectomy in order to distinguish this appearance from pathologic changes that may be seen in the presence of infection or recurrent tumors. METHODS: We reviewed 159 MR scans from 95 patients with hippocampal sclerosis or gliosis who underwent temporal lobectomy for treatment of intractable epilepsy. Choroid plexus location and size were assessed on contrast-enhanced T1-weighted images. RESULTS: After temporal lobectomy, the choroid plexus enlarged and sagged into the resection site. Increase in the size of the choroid plexus occurred in 58% of cases overall. The degree of enhancement also increased after surgery, sometimes resulting in a nodular pattern of enhancement. The changes were most marked during the 1st week after temporal lobectomy, and showed an enlarged, markedly enhancing choroid plexus on 86% of the scans. CONCLUSION: Postoperative changes of the choroid plexus after temporal lobectomy include sagging into the resection site, an increased size, and an increased degree of enhancement. Normal postoperative morphologic characteristics may mimic neoplastic enhancement pattern. Familiarity with this appearance is important to avoid a pitfall in diagnosis of recurrent postoperative temporal lobe neoplasms.

Effects of cell volume fraction changes on apparent diffusion in human cells.

Diffusion-weighted imaging was used to study the relationship between apparent diffusion coefficient (ADC) and cell volume fraction in cell suspensions and packed arrays. Cell volume fraction was varied by changing extracellular fluid osmolarity (for human glial cells) and by changing cell density (for human glial and red blood cells). In packed arrays of glial cells, ADC increased 10% when cells shrank and decreased 13% when cells swelled. ADC decreased 34% as cell density increased from 0 to 72%. In erythrocyte suspensions, ADC decreased 90% as the cell density increased from 0 to 89%. These results agree with theoretical predictions.

Cerebrospinal fluid cleft with cortical dimple: MR imaging marker for focal cortical dysgenesis.

PURPOSE: To determine whether the magnetic resonance (MR) imaging feature of a cerebrospinal fluid (CSF) cleft and cortical dimple can be used as a marker for cortical dysgenesis. MATERIALS AND METHODS: MR images in 875 patients with epilepsy were evaluated for the following features of focal cortical dysgenesis: cortical thickening, indistinct junction between gray and white matter, macrogyria, and abnormal sulcal pattern. Images with these features were reevaluated to determine the relationship between the CSF cleft-cortical dimple complex and focal cortical dysgenesis and its contribution to diagnosis. The cleft-dimple complex was defined as a prominent CSF space associated with a region of cortical volume loss adjacent to the dysgenesis. RESULTS: Seventy-one patients had cortical dysgenesis, including 27 with cellular proliferation abnormalities, 18 with migration abnormalities, 25 with cortical organization abnormalities, and one with miscellaneous anomalies. Histologic correlation was available in 20 patients. There was an associated cortical dimple in 29 (41%) patients. This association was strongest in patients with cortical organization abnormalities: It was present in 22 of the 25 (88%) patients. In 12 (48%) patients with abnormalities of cortical organization, the CSF cleft was easier to detect than other features of cortical dysgenesis or contributed directly to the MR imaging diagnosis. CONCLUSION: The cleft-dimple complex is a marker for subtle cases of focal cortical dysgenesis and may be due to maldevelopment. In patients with seizures, the presence of a cleft-dimple complex should alert the physician to scrutinize adjacent regions for developmental anomalies.

MR imaging and histologic features of subinsular bright spots on T2-weighted MR images: Virchow-Robin spaces of the extreme capsule and insular cortex.

PURPOSE: To determine the cause and frequency of high-signal-intensity foci detected in the insular cortex and extreme capsule on thin-section, high-spatial-resolution, coronal, T2-weighted magnetic resonance (MR) images. MATERIALS AND METHODS: The authors assessed high-signal-intensity areas in the insular cortex and extreme capsule on coronal MR images obtained in 56 patients with seizure and five control subjects. Images were obtained with thin-section, high-spatial-resolution, T2-weighted, fast spin-echo; three-dimensional, spoiled gradient-recalled-echo; and fluid-attenuated inversion-recovery sequences. In two formalin-fixed brain specimens, MR imaging findings were correlated with gross anatomic and histologic findings. RESULTS: Subinsular bright spots were found in 53 of the 56 (95%) patients (96 of 112 [86%] hemispheres) and all five control subjects. The spots were elliptical in 30 patients, round in 14 patients, linear in 22 patients, and dotlike in seven patients and often had a featherlike configuration. The spots were isointense to cerebrospinal fluid on T2-weighted, fast SE images and were located in the anterior extreme capsule white matter and insular cortex. MR imaging of brain specimens revealed bilateral elliptical areas of high signal intensity that corresponded to small multiple cavities at gross anatomic inspection. At microscopic examination, these cavities were perivascular spaces of mostly arteriolar origin. CONCLUSION: High-signal-intensity subinsular foci at MR imaging are due to enlarged perivascular spaces. In most cases, these foci can be visualized on thin-section, high-spatial-resolution, coronal T2-weighted images; they should not be mistaken for pathologic conditions when they occur unilaterally.

MR imaging of implanted depth and subdural electrodes: is it safe?

This study evaluates the safety of imaging chronic epilepsy patients with intracranial depth and subdural electrodes by magnetic resonance (MR). To identify an epileptogenic focus, the precise location of the electrode contacts is necessary, and MR can provide this information. However, many neurosurgeons and neuroradiologists are hesitant to scan patients by MR with these implanted, metallic electrodes for fear of electrode displacement, current induction or heating secondary to the strong magnetic field. In the present study, the subdural electrodes were made of stainless steel with either stainless steel or platinum contacts. The depth electrodes were made of either platinum or a nickel-chromium alloy (nichrome). We reviewed 98 cases in which patients with implanted depth electrodes, subdural electrodes, or both underwent MR scanning. A total of 143 depth electrodes, 688 subdural strips, and 38 subdural grids were implanted in the 98 procedures. MR scanning was performed on a 1.5-T unit and consisted of T1, T2, and/or spoiled gradient echo pulse sequences. There were no documented complications related to the MR scans. Based on this study and a review of the literature, we feel that MR imaging can safely localize intracranial electrodes.

Surgical outcome in occipital lobe epilepsy: implications for pathophysiology.

Medically refractory occipital lobe epilepsies are increasingly treated with surgery, but outcome and its relationship to etiology, pathological substrate, occipital lobe location, surgical approach, and electroclinical features have not been systematically investigated in a substantial group of patients. Thirty-five patients who underwent surgery for intractable occipital lobe seizures were retrospectively evaluated. Outcome and occipital lobe location were analyzed with respect to surgical procedure, pathology, clinical seizure characteristics, seizure onset and termination locations, and localization of interictal spikes. Most patients had developmental abnormalities (14) or tumors (13, all gliomas). Developmental abnormalities consisted of focal cortical dysplasia (5), heterotopia (2), hamartoma (3), cortical duplication (1), polymicrogyria (1), Sturge-Weber syndrome (1), and tuberous sclerosis (1). There was 1 patient with a vascular abnormality, 1 with chronic inflammatory changes, 4 with gliosis, 1 with cerebral ossification, and 1 with normal pathology. Developmental abnormalities had significantly worse outcome (45% excellent/good) than tumors (85% excellent/good). In the developmental group, low-grade focal cortical dysplasias had better outcome than heterotopia and hamartoma regardless of type of surgical procedure. Pathological groups did not significantly differ with respect to location within the occipital lobe (overall medial [50%] or lateral [38%]); clinical seizure characteristics referable to specific lobe (occipital [14%], temporal [34%], frontal [23%], more than one type [29%]); electroencephalographic localization (to occipital [17%], temporal [27%], or other/multifocal locations [56%]); or intracranial ictal onset or termination location. Electroclinical variables were also unrelated to the occipital lobe location of abnormality. Surgical outcome was not predicted by surgical approach (lesion excision with margins or lobectomy). The main pathological substrates of uncontrolled occipital lobe epilepsy are gliomas and developmental abnormalities. Whereas resection of occipital lobe tumors associated with chronic epilepsy produces nearly uniform seizure control, outcome after resection of occipital lobe developmental abnormalities is less uniform.

Postoperative changes in the brain: MR imaging findings in patients without neoplasms.

PURPOSE: To determine the time course of postoperative changes in the brain by using magnetic resonance (MR) imaging. MATERIALS AND METHODS: Contrast material-enhanced MR images (32 preoperative, 161 postoperative) in 95 patients who underwent temporal lobectomy for treatment of nonneoplastic epilepsy were evaluated for enhancement pattern at the surgical margin and for the presence of dural enhancement, fluid, and air. Images were obtained 9 hours to 5.6 years after surgery. RESULTS: Surgical margin contrast enhancement was present and increased from 20 hours to 29 days after surgery. The predominant early (days 0-5) pattern was thin linear enhancement (16 of 18 [89%] images). The later (days 6-29) pattern was thick linear or nodular (tumorlike appearance) (11 of 18 [61%] images). Surgical margin enhancement was absent after day 30, although few data were collected 1-5 months after surgery. Dural enhancement was seen on 96 of 97 (99%) postoperative MR images. Extraaxial fluid was seen on all 46 images obtained during the first 30 days and was absent by day 51. Intracranial air was present on 31 of 35 (89%) images obtained during the first 4 days but was absent after day 5. CONCLUSION: Knowledge of the evolution of contrast enhancement patterns on postoperative MR images of the brain can help in differentiating benign from neoplastic changes.

Qualitative MR imaging of refractory temporal lobe epilepsy requiring surgery: correlation with pathology and seizure outcome after surgery.

OBJECTIVE: The purpose of this study was to compare MR imaging findings with histologic findings of either hippocampal sclerosis or gliosis in patients with intractable temporal lobe epilepsy requiring surgery and to correlate MR imaging findings with seizure outcome after surgery and with clinical parameters such as febrile seizure history. MATERIALS AND METHODS: A retrospective study of MR scans of 66 patients with medically refractory temporal lobe epilepsy requiring surgery was performed. Qualitative diagnosis was done by visual inspection of MR images. MR imaging findings of hippocampal atrophy, signal intensity changes, and segmental findings were correlated with histopathology and with neuronal density. The final MR imaging diagnosis was also correlated with seizure outcome after surgery and with febrile seizure history. RESULTS: Histologic findings consisted of hippocampal sclerosis in 55 patients and nonspecific gliosis in 11 patients. Two variables, MR imaging findings of hippocampal sclerosis (hippocampal atrophy or signal intensity change) and a febrile seizure history, were significantly associated with hippocampal sclerosis. MR images of nonspecific gliosis usually showed normal findings, although some cases showed mild hippocampal atrophy. When comparing MR imaging findings with histology, our observers achieved sensitivities of 87-98% and specificities of 45-100%. In patients with successful outcomes after surgery, sensitivity ranged from 85% to 98% for MR imaging findings suggesting hippocampal sclerosis, specificity ranged from 17% to 85%, and positive predictive values ranged from 82% to 90%. CONCLUSION: Qualitative visual analysis of MR images correlates well with histologic findings, febrile seizure history, and seizure outcome after surgery. MR imaging findings and febrile seizure history help differentiate between hippocampal sclerosis and nonspecific gliosis, two similar clinical conditions associated with temporal lobe epilepsy that often have different outcomes after surgery.

A systematic approach for interpreting MR images of the seizure patient.

A systematic approach needs to be used to review MR scans in epilepsy patients to avoid the common pitfalls engendered by the subtle nature of many epileptogenic lesions. One should always evaluate the hippocampus regardless of other MR findings to avoid missing dual abnormalities. False-positive and false-negative diagnosis of hippocampal sclerosis can be avoided by evaluating the hippocampus after correcting for head rotation [by assessing the internal auditory canals and atria). Periventricular heterotopia can be successfully diagnosed by systematically studying the periventricular regions, especially those adjacent to the atria of the lateral ventricles. Gray matter lateral to the ventricles (excluding the caudate nucleus) is always an abnormal finding. Sulcal and cortical morphologic abnormalities are particularly difficult to diagnose unless a high index of suspicion is maintained. Cortical thickening is indicative of a developmental anomaly and should be screened in an organized manner. Because epilepsy is generally a cortical process, one must search for subtle cortical abnormalities, including focal atrophic abnormalities and lesions without mass effect. Diligence will offer its own rewards.

Focal cortical dysplasia of Taylor, balloon cell subtype: MR differentiation from low-grade tumors.

PURPOSE: To test the hypothesis that focal cortical dysplasia of Taylor (FCDT) can be distinguished from low-grade tumors by means of clinical and MR findings. METHODS: We examined 10 clinical and 19 MR imaging variables in patients who underwent surgery for intractable epilepsy over an 8-year period. The 54 patients with low-grade glial neoplasms were compared with the eight patients who had balloon cell FCDT. RESULTS: Statistically significant differences were seen with respect to eight of the MR variables and none of the clinical variables. MR findings suggesting dysplasia rather than tumor included the presence of gray matter thickening associated with a homogeneous hyperintense signal in the subcortical white matter that tapers as it extends to the lateral ventricle. A frontal lobe location favors dysplasia, while a temporal lobe (especially medial temporal lobe) location is more suggestive of a neoplasm. CONCLUSION: Several MR features help distinguish balloon cell FCDT from neoplasms, especially cortical thickening and a tapered signal to the ventricle. This distinction is important for surgical planning, as the decision to operate and the extent of surgical resection often depend on the presence or absence of neoplastic tissue.

Predicting reading performance from neuroimaging profiles: the cerebral basis of phonological effects in printed word identification.

This study linked 2 experimental paradigms for the analytic study of reading that heretofore have been used separately. Measures on a lexical decision task designed to isolate phonological effects in the identification of printed words were examined in young adults. The results were related to previously obtained measures of brain activation patterns for these participants derived from functional magnetic resonance imaging (fMRI). The fMRI measures were taken as the participants performed tasks that were designed to isolate orthographic, phonological, and lexical-semantic processes in reading. Individual differences in the magnitude of phonological effects in word recognition, as indicated by spelling-to-sound regularity effects on lexical decision latencies and by sensitivity to stimulus length effects, were strongly related to differences in the degree of hemispheric lateralization in 2 cortical regions.

Positron emission tomography measurement of cerebral metabolic correlates of tryptophan depletion-induced depressive relapse.

BACKGROUND: Short-term depletion of plasma tryptophan has been shown to result in depressive relapse in patients with remission of major depression. Positron emission tomography and single photon emission computed tomography studies implicated the dorsolateral prefrontal cortex, orbitofrontal cortex, thalamus, and caudate nucleus in the pathogenesis of depression. The purpose of this study was to measure cerebral metabolic correlates of tryptophan depletion-induced depressive relapse. METHODS: Patients diagnosed as having major depression (N = 21) who clinically improved with serotonin reuptake inhibitors underwent 2 test days involving tryptophan depletion or placebo, followed 6 hours later by positron emission tomography scanning with fludeoxy-glucose F18. Brain metabolism was compared in patients with (n = 7) and without (n = 14) a tryptophan depletion-induced depressive relapse. RESULTS: Tryptophan depletion resulted in a decrease in brain metabolism in the middle frontal gyrus (dorsolateral prefrontal cortex), thalamus, and orbitofrontal cortex in patients with a depletion-induced depressive relapse (but not in patients without depletion-induced relapse). Decreased brain metabolism in these regions correlated with increased depressive symptoms. Baseline metabolism was increased in prefrontal and limbic regions in relapse-prone patients. CONCLUSION: Specific brain regions, including the middle frontal gyrus, thalamus, and orbitofrontal cortex, may mediate the symptoms of patients with major depression.

Embryology of the human fetal hippocampus: MR imaging, anatomy, and histology.

PURPOSE: To identify changes in the embryology of the hippocampus responsible for its adult anatomy. METHODS: Ten human fetal specimens ranging from 13 to 24 weeks' gestational age were examined with MR imaging. Dissections and histologic sections of 10 different specimens of similar ages were compared with MR imaging findings. RESULTS: At 13 to 14 weeks' gestation, the unfolded hippocampus, on the medial surface of the temporal lobe, surrounds a widely open hippocampal sulcus (hippocampal fissure). At 15 to 16 weeks, the dentate gyrus and cornu ammonis have started to infold. The hippocampal sulcus remains open. The parahippocampal gyrus is larger and more medially positioned. The CA1, CA2, and CA3 fields of the cornu ammonis are arranged linearly. The dentate gyrus has a narrow U shape. By 18 to 20 weeks, the hippocampus begins to resemble the adult hippocampus. The dentate gyrus and cornu ammonis have folded into the temporal lobe. The hippocampus and subiculum approximate each other across a narrow hippocampal sulcus. The CA1-3 fields form an arc and the CA4 field has increased in size within the widened arch of the dentate gyrus. CONCLUSION: MR imaging of fetuses provides a developmental basis for understanding hippocampal anatomy.